JT> Reflections garble the data enough to make it useless for
JT> everyone let alone the transmitter trying to detect a collision
JT> or not.

JB> IIRC Signals travel in coax at arround 200 Mm/s (2/3
JB> lightspeed) so a 50 ns pulse (10Mbps ethernet) is about 10m
JB> long

JB> That short.

JB> Hmm, maybe I should repeat that experiment with a shorter
JB> cable.

 That happens whether terminated or not. The whole train of pulses
shift.

 What happens to an unterminated pulsee, is resonance, an overshoot
on the edges, and ringing. Ten-metres is 120MHz for quarter-wave
resonance (about 100MHz in thin coax), or a risetime of 2.5 nS. I
would not expect that a 2.5nS ring would matter a rat's arse... and a
longer run would damp the ring with lossy cable anyway.

 SHIT!!

 I've just realised that I've got it backwards. I've been multiplying
quarter-wave resonance instead of dividing. A quarter wave 10-metres
is not 150MHZ... it's 8MHZ! The ringing edge would therefore be
spot-on resonant with the pulse. Instead of a nice square pulse,
you'll get damped sinewaves.

 Sorry...

Regards,
Bob

  

--- BQWK Alpha 0.5